Polling increasingly difficult numbers game

What you expected to happen on election night depended entirely on who you asked and when you asked.

The NDP was within reach of forming government; or there was a dead heat; or the Conservatives were pulling ahead; or the Liberals were surging.

The numbers game is an increasingly difficult one for pollsters who try and gauge voter intentions.

So tough, in fact, that the granddaddy of political polling – Gallup – called it quits earlier this month, announcing it will not conduct any horse-race polling for the U.S. presidential campaign for the first time in eight decades.

It was embarrassingly off on its 2012 election predictions and publicly called to the carpet by Barack Obama’s campaign team.

“Gallup was one of the grand old names in polling but Gallup has had particular problems,” says Barry Kay, an associate professor of political science at Wilfrid Laurier University and number cruncher for the Laurier Institute for the Study of Public Opinion and Policy.

“Their reputation, in recent American elections, had become very bad and they decided … that discretion was the better part of valor.”

In Canada, the 2013 British Columbia election was a low point for pollsters. Even the victorious B.C. Liberals seemed stunned on election night when the ballots started coming in.

The Wildrose results in Alberta the year before offered political prognosticators no bragging rights and Quebec Liberals were relieved, to say the least, that they were not decimated as predicted in the provincial election that fall.

“The state of polling has never been worse nor ever been in greater demand,” Kay tells Yahoo Canada News.

Technological advances such as cellphones, robopolling, telemarketing and caller ID have created challenges.

A generation ago, the participation rate was about 75 per cent; today, less than one in 10 voters is reached. Gender and generational disparities abound, he says.

But the demand for polls have never been higher. Indeed, polls were everywhere during this lengthy campaign, lurching week to week from one party to another.

But that was a reflection more of an undecided electorate than a failure of polling, says Frank Graves, president of EKOS Research Associates.

“To be honest, we feel pretty comfortable with our final poll and the analysis and recommendations or predictions that we made,” he tells Yahoo Canada News. “We basically called what was going to happen.”

On the eve of the election, EKOS and others predicted a Liberal win and possibly a majority. They said Quebec and B.C. were areas of uncertainty.

The frenetic changes throughout the campaign were an accurate reflection of a shifting electorate, he says. The ups and downs of party fortunes reflected the issues that arose during the campaign.

“We’re very confident that our polls were measuring real changes throughout the campaign,” Graves says.

There are hurdles to modern polling, he admits, but in this case the result was accurate.

Canada has a different regulatory structure and culture that means the proliferation of cellphones is not as acute as in the United States, he says.

“It’s a manageable problem,” Graves says. “Overall, we were very close to … to what we predicted.”

Kay is less impressed with the polls or the seat projections he and other poll aggregators used to make seat projections.

Though the institute predicted a Liberal government, it was a minority. The inaccuracies in Quebec, in particular, were “most egregious.”

Those predictions were based on an aggregation of as many as eight different poll results.

“Garbage in, garbage out,” Kay says. “If the polling data we’re using is no good, the projections is going to be no good.”

Despite their problems, political polls and projections are not likely to disappear any time soon, he says.

“Next election I expect we’ll have more polls than this time because there’s lots of media outlets that find putting forward polls… increases their viewership and there’s lots of pollsters that want the advertising associated with it,” Kay says.